blob: 396337301c774a2bed86b5d548600503a0f1cd38 [file] [log] [blame]
// Copyright 2014 The Crashpad Authors. All rights reserved.
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
#include "util/mach/mach_message_server.h"
#include <string.h>
#include <limits>
#include "base/logging.h"
#include "base/mac/mach_logging.h"
#include "base/mac/scoped_mach_vm.h"
#include "util/mach/mach_message.h"
namespace crashpad {
namespace {
//! \brief Manages a dynamically-allocated buffer to be used for Mach messaging.
class MachMessageBuffer {
public:
MachMessageBuffer() : vm_() {}
~MachMessageBuffer() {}
//! \return A pointer to the buffer.
mach_msg_header_t* Header() const {
return reinterpret_cast<mach_msg_header_t*>(vm_.address());
}
//! \brief Ensures that this object has a buffer of exactly \a size bytes
//! available.
//!
//! If the existing buffer is a different size, it will be reallocated without
//! copying any of the old buffer’s contents to the new buffer. The contents
//! of the buffer are unspecified after this call, even if no reallocation is
//! performed.
kern_return_t Reallocate(vm_size_t size) {
// This test uses == instead of > so that a large reallocation to receive a
// large message doesn’t cause permanent memory bloat for the duration of
// a MachMessageServer::Run() loop.
if (size != vm_.size()) {
// reset() first, so that two allocations don’t exist simultaneously.
vm_.reset();
if (size) {
vm_address_t address;
kern_return_t kr =
vm_allocate(mach_task_self(),
&address,
size,
VM_FLAGS_ANYWHERE | VM_MAKE_TAG(VM_MEMORY_MACH_MSG));
if (kr != KERN_SUCCESS) {
return kr;
}
vm_.reset(address, size);
}
}
#if !defined(NDEBUG)
// Regardless of whether the allocation was changed, scribble over the
// memory to make sure that nothing relies on zero-initialization or stale
// contents.
memset(Header(), 0x66, size);
#endif
return KERN_SUCCESS;
}
private:
base::mac::ScopedMachVM vm_;
DISALLOW_COPY_AND_ASSIGN(MachMessageBuffer);
};
// Wraps MachMessageWithDeadline(), using a MachMessageBuffer argument which
// will be resized to |receive_size| (after being page-rounded). MACH_RCV_MSG
// is always combined into |options|.
mach_msg_return_t MachMessageAllocateReceive(MachMessageBuffer* request,
mach_msg_option_t options,
mach_msg_size_t receive_size,
mach_port_name_t receive_port,
MachMessageDeadline deadline,
mach_port_name_t notify_port,
bool run_even_if_expired) {
mach_msg_size_t request_alloc = round_page(receive_size);
kern_return_t kr = request->Reallocate(request_alloc);
if (kr != KERN_SUCCESS) {
return kr;
}
return MachMessageWithDeadline(request->Header(),
options | MACH_RCV_MSG,
receive_size,
receive_port,
deadline,
notify_port,
run_even_if_expired);
}
} // namespace
// This method implements a server similar to 10.9.4
// xnu-2422.110.17/libsyscall/mach/mach_msg.c mach_msg_server_once(). The server
// callback function and |max_size| parameter have been replaced with a C++
// interface. The |persistent| parameter has been added, allowing this method to
// serve as a stand-in for mach_msg_server(). The |timeout_ms| parameter has
// been added, allowing this function to not block indefinitely.
//
// static
mach_msg_return_t MachMessageServer::Run(Interface* interface,
mach_port_t receive_port,
mach_msg_options_t options,
Persistent persistent,
ReceiveLarge receive_large,
mach_msg_timeout_t timeout_ms) {
options &= ~(MACH_RCV_MSG | MACH_SEND_MSG);
const MachMessageDeadline deadline =
MachMessageDeadlineFromTimeout(timeout_ms);
if (receive_large == kReceiveLargeResize) {
options |= MACH_RCV_LARGE;
} else {
options &= ~MACH_RCV_LARGE;
}
const mach_msg_size_t trailer_alloc = REQUESTED_TRAILER_SIZE(options);
const mach_msg_size_t expected_receive_size =
round_msg(interface->MachMessageServerRequestSize()) + trailer_alloc;
const mach_msg_size_t request_size = (receive_large == kReceiveLargeResize)
? round_page(expected_receive_size)
: expected_receive_size;
DCHECK_GE(request_size, sizeof(mach_msg_empty_rcv_t));
// mach_msg_server() and mach_msg_server_once() would consider whether
// |options| contains MACH_SEND_TRAILER and include MAX_TRAILER_SIZE in this
// computation if it does, but that option is ineffective on macOS.
const mach_msg_size_t reply_size = interface->MachMessageServerReplySize();
DCHECK_GE(reply_size, sizeof(mach_msg_empty_send_t));
const mach_msg_size_t reply_alloc = round_page(reply_size);
MachMessageBuffer request;
MachMessageBuffer reply;
bool received_any_request = false;
bool retry;
kern_return_t kr;
do {
retry = false;
kr = MachMessageAllocateReceive(&request,
options,
request_size,
receive_port,
deadline,
MACH_PORT_NULL,
!received_any_request);
if (kr == MACH_RCV_TOO_LARGE) {
switch (receive_large) {
case kReceiveLargeError:
break;
case kReceiveLargeIgnore:
// Try again, even in one-shot mode. The caller is expecting this
// method to take action on the first message in the queue, and has
// indicated that they want large messages to be ignored. The
// alternatives, which might involve returning MACH_MSG_SUCCESS,
// MACH_RCV_TIMED_OUT, or MACH_RCV_TOO_LARGE to a caller that
// specified one-shot behavior, all seem less correct than retrying.
MACH_LOG(WARNING, kr) << "mach_msg: ignoring large message";
retry = true;
continue;
case kReceiveLargeResize: {
mach_msg_size_t this_request_size = round_page(
round_msg(request.Header()->msgh_size) + trailer_alloc);
DCHECK_GT(this_request_size, request_size);
kr = MachMessageAllocateReceive(&request,
options & ~MACH_RCV_LARGE,
this_request_size,
receive_port,
deadline,
MACH_PORT_NULL,
!received_any_request);
break;
}
}
}
if (kr != MACH_MSG_SUCCESS) {
return kr;
}
received_any_request = true;
kr = reply.Reallocate(reply_alloc);
if (kr != KERN_SUCCESS) {
return kr;
}
mach_msg_header_t* request_header = request.Header();
mach_msg_header_t* reply_header = reply.Header();
bool destroy_complex_request = false;
interface->MachMessageServerFunction(
request_header, reply_header, &destroy_complex_request);
if (!(reply_header->msgh_bits & MACH_MSGH_BITS_COMPLEX)) {
// This only works if the reply message is not complex, because otherwise,
// the location of the RetCode field is not known. It should be possible
// to locate the RetCode field by looking beyond the descriptors in a
// complex reply message, but this is not currently done. This behavior
// has not proven itself necessary in practice, and it’s not done by
// mach_msg_server() or mach_msg_server_once() either.
mig_reply_error_t* reply_mig =
reinterpret_cast<mig_reply_error_t*>(reply_header);
if (reply_mig->RetCode == MIG_NO_REPLY) {
reply_header->msgh_remote_port = MACH_PORT_NULL;
} else if (reply_mig->RetCode != KERN_SUCCESS &&
request_header->msgh_bits & MACH_MSGH_BITS_COMPLEX) {
destroy_complex_request = true;
}
}
if (destroy_complex_request &&
request_header->msgh_bits & MACH_MSGH_BITS_COMPLEX) {
request_header->msgh_remote_port = MACH_PORT_NULL;
mach_msg_destroy(request_header);
}
if (reply_header->msgh_remote_port != MACH_PORT_NULL) {
// Avoid blocking indefinitely. This duplicates the logic in 10.9.5
// xnu-2422.115.4/libsyscall/mach/mach_msg.c mach_msg_server_once(),
// although the special provision for sending to a send-once right is not
// made, because kernel keeps sends to a send-once right on the fast path
// without considering the user-specified timeout. See 10.9.5
// xnu-2422.115.4/osfmk/ipc/ipc_mqueue.c ipc_mqueue_send().
const MachMessageDeadline send_deadline =
deadline == kMachMessageDeadlineWaitIndefinitely
? kMachMessageDeadlineNonblocking
: deadline;
kr = MachMessageWithDeadline(reply_header,
options | MACH_SEND_MSG,
0,
MACH_PORT_NULL,
send_deadline,
MACH_PORT_NULL,
true);
if (kr != MACH_MSG_SUCCESS) {
if (kr == MACH_SEND_INVALID_DEST ||
kr == MACH_SEND_TIMED_OUT ||
kr == MACH_SEND_INTERRUPTED) {
mach_msg_destroy(reply_header);
}
return kr;
}
}
} while (persistent == kPersistent || retry);
return kr;
}
} // namespace crashpad